To develop efficient experimental designs for selecting the best of several experimental treatments. Data from controlled laboratory or clinical trials are assumed to occur sequentially in time where the individual responses are dichotomous. Sequential regions have been developed for deciding when to terminate the experiment and which treatment to select as most effective. Improved likelihood stopping procedures have been developed which reduce the total number of patients in the clinical trial. Techniques for the inclusion of concomitant information such as patient age differences, differing clinical centers, etc., have been produced.